Good question. My guess would be that transforms along ridges do not really help decide if there is whole mantle or just UM convection. The main force that influences the lithosphere is probably slab pull and so transforms may have more to do with THAT rather than the depth of convection per se.

I dont think that the observation of ridge segments offset by fracture zones really supports or refutes either of these two hypotheses for plate motion mechanisms. The real significance of fracture zones is that they probably accommodate strain because a continuous, non-segmented spreading ridge is not geometrically stable on a sphere. We describe the motion of one tectonic plate relative to another as a rotation around an Euler pole, and arcuate fracture zones accommodate this rotation.

Concerning the hypotheses described in your question, I think that the dominant current opinion is that the weight of sinking subducting slabs drives spreading at ridges, as opposed to large total-mantle convection cells dragging at the base of oceanic crust. You should check out this review paper by Warren Hamilton (2007) for a rather flagrant (and convincing) address to this topic. He provides many lines of evidence that I wont take the time to discuss here. I think that there is profound evidence for upper mantle convection, and this may enhance the motion of tectonic plates, but I think that it is probably a feedback process.

Sorry if deviated from your question, but it seems as though you are trying to address the competing models of whole-mantle convection driving plate tectonics versus slab pull as the dominant mechanism.

Hamilton, W.B. (2007). Driving mechanism and 3-D circulation of plate tectonics. GSA special paper.

Good question! Given the scale of ridge offset, I would be surprised if it's that deep into the mantle, but I don't know. Given the paucity of description of what I was able to find with a brief Google search, I wouldn't be surprised if the answer isn't known. However, one of the first links I found was Dr. Macdonald at UCSB - you might try sending an email. Good luck!